JPWO2008032753A1 - Polishing apparatus and polishing method - Google Patents

Abstract

The present invention relates to a polishing apparatus and a polishing method for polishing and flattening a substrate such as a semiconductor wafer. A polishing apparatus according to the present invention includes a polishing table (10) having a polishing surface and a top ring (pressing the substrate against the polishing table by independently applying a pressing force to the first plurality of regions on the substrate. 14), a sensor (50) for detecting the state of the film at a plurality of measurement points, and a monitoring device (53) for generating a monitoring signal for each of the second plurality of regions on the substrate from the output signal of the sensor, The storage unit storing a plurality of reference signals indicating the relationship between the reference value of the monitoring signal and the polishing time, and the monitoring signal corresponding to each of the second plurality of regions converges to any one of the plurality of reference signals And a control unit for operating the pressing force on the first plurality of regions.

Description

  The present invention relates to a polishing apparatus and a polishing method, and more particularly to a polishing apparatus and a polishing method for polishing and flattening a substrate such as a semiconductor wafer.

As a polishing apparatus for polishing and flattening a substrate such as a semiconductor wafer, an apparatus capable of independently adjusting the pressures of a plurality of chambers in a top ring is known. In this polishing apparatus, for example, a sensor measures a physical quantity related to the film thickness on the substrate, and a monitoring signal is generated based on the physical quantity. Before polishing the substrate, a reference signal indicating the relationship between the monitoring signal and time is prepared in advance. During polishing, the top ring is used so that the monitoring signal at each measurement point on the substrate converges to the reference signal. The pressing force is adjusted. Thereby, a uniform residual film thickness is realized in the substrate surface (see, for example, WO 2005/123335).
However, the conventional polishing apparatus has a problem that the sensor signal value acquired in a certain area of the substrate may be significantly different from that in other areas, and the sensor cannot correctly evaluate the film thickness. One reason for this is a decrease in signal due to the effective measurement range of the sensor. The effective measurement range of the sensor necessarily has a certain size. For this reason, when it is going to measure the vicinity of the peripheral part of a wafer, a part of effective measurement range of a sensor will protrude from the to-be-polished surface of a wafer, and a sensor cannot acquire an exact signal. In such a case, the control can be performed by excluding the measurement points where the accurate signal cannot be obtained. However, this method is particularly effective when the film thickness uniformity at the peripheral edge of the wafer is important. Cannot be taken.
Another cause is the influence of the metal and magnetic material in the top ring. That is, when a conductive metal part such as SUS or a magnetic material is used for the top ring, the signal value of the sensor locally changes due to the influence.

The present invention has been made in view of the above-described problems of the prior art, and an object thereof is to provide a polishing apparatus and a polishing method capable of accurately controlling a film thickness profile after polishing a substrate.
In order to achieve the above object, one embodiment of the present invention is a polishing apparatus for polishing a substrate having a film formed on a surface, the polishing table having a polishing surface, and a first plurality of regions on the substrate A top ring that presses the substrate against the polishing table by applying a pressing force independently to the polishing table, a sensor that detects the state of the film at a plurality of measurement points, and an output signal of the sensor, A monitoring device that generates a monitoring signal for each of the plurality of regions, a storage unit that stores a plurality of reference signals indicating a relationship between a reference value of the monitoring signal and a polishing time, and a second plurality of regions A control unit for operating a pressing force on the first plurality of regions so that the monitoring signal corresponding to each converges to any one of the plurality of reference signals. And butterflies.
In a preferred aspect of the present invention, one of the second plurality of regions is a region including a peripheral portion of the substrate, and one of the plurality of reference signals is a reference for a region including the peripheral portion of the substrate. It is a signal.
In a preferred aspect of the present invention, the plurality of reference signals are provided corresponding to the second plurality of regions, respectively.
In a preferred aspect of the present invention, a signal value of the monitoring signal and a signal value of the reference signal are converted into a value related to a polishing time based on the reference signal, and a new monitoring signal and a new reference signal are generated. It is characterized by doing.
In a preferred aspect of the present invention, a value obtained by averaging the new monitoring signals in the second plurality of regions is obtained at an arbitrary time of the polishing process, and the new reference signal at the time is the averaged value. The new reference signal after the time is moved in parallel with respect to the time axis so as to coincide with.
In a preferred aspect of the present invention, the plurality of reference signals correspond to the same film thickness at the same time.
In a preferred aspect of the present invention, the plurality of reference signals correspond to film thicknesses reflecting a predetermined film thickness difference set between the second plurality of regions at the same time point.
In a preferred aspect of the present invention, the control period of the control unit is not less than 1 second and not more than 10 seconds.
In a preferred aspect of the present invention, the sensor is an eddy current sensor.
In a preferred aspect of the present invention, the control unit detects a polishing end point based on a monitoring signal generated by the monitoring device.
Another aspect of the present invention is a polishing method in which a substrate is pressed against a polishing table by applying an independent pressing force to the first plurality of regions on the substrate, and the film thickness on the substrate is increased. Define a plurality of reference signals indicating the relationship between the reference value of the related monitoring signal and the polishing time, detect the state of the film on the substrate at a plurality of measurement points using a sensor, and from the sensor output signal, A monitoring signal is generated for each of the second plurality of regions, and the monitoring signal corresponding to each of the second plurality of regions converges to any one of the plurality of reference signals. It is characterized by operating a pressing force for a plurality of one region.
In a preferred aspect of the present invention, a reference substrate of the same type as the substrate to be polished is prepared, the thickness of the reference substrate is measured, the reference substrate is polished, and the film on the reference substrate at a plurality of measurement points is measured. A state is detected by the sensor, and a monitoring signal in the first region and the second region selected from the second plurality of regions is generated from the output signal of the sensor, and the first region and the second region The polishing is stopped when the film to be polished in the region is completely removed, and the average polishing rate of the first region and the second region is obtained. The average polishing rate of the second region is the first polishing rate. The monitoring signal of the second region is stretched or reduced along the time axis so as to match the average polishing rate of the region, and the initial film thickness of the second region is equal to the initial film thickness of the first region. Find the polishing time required to match The translated signal of the stretched or shrunk second region is translated along the time axis for the determined polishing time, and the translated monitoring signal is used as a reference signal for the second region. The plurality of reference signals are defined.
According to the present invention, since a plurality of reference signals are provided for a plurality of regions on the substrate, a uniform film thickness can be obtained in the entire region of the substrate. In addition, since it is not necessary to bring the sensor close to the surface to be polished of the substrate in order to reduce the effective measurement range of the sensor, it is possible to use a normal polishing pad having no through hole or back surface depression.

FIG. 1 is a schematic diagram showing an overall configuration of a polishing apparatus according to an embodiment of the present invention.
FIG. 2 is a schematic view showing a cross section of the top ring shown in FIG.
FIG. 3 is a plan view showing the relationship between the polishing table and the wafer.
FIG. 4 is a diagram showing a trajectory that the sensor scans on the wafer.
FIG. 5 is a plan view showing an example of selecting measurement points to be monitored by the monitoring device from the measurement points on the wafer shown in FIG.
FIG. 6 is a diagram illustrating an effective measurement range of the sensor at each measurement point.
FIG. 7 is a graph showing signal values in each region on the wafer.
FIG. 8 is a flowchart showing a flow of creating a reference signal for each region based on a monitoring signal when the reference wafer is polished.
9A and 9B are schematic diagrams showing examples of film thickness distribution.
FIG. 10 is an example of a monitoring signal when the reference wafer is polished.
FIG. 11 is a diagram for explaining the scaling with respect to the time axis of the monitoring signal.
FIG. 12 is a diagram for explaining a method of further translating the monitoring signal scaled along the time axis along the time axis.
FIG. 13 is a graph for explaining an example of a method for converting the reference signal and the monitoring signal.
FIG. 14 is a graph for explaining an example of a method of applying a reference signal.
FIG. 15 is a graph for explaining another example of the application method of the reference signal.
FIG. 16 is a graph for explaining another example of the application method of the reference signal.
FIG. 17 is a graph showing the radial film thickness distribution before and after polishing when a reference signal is generated and polishing is performed.
FIG. 18 is a graph showing the transition of the monitoring signal in uncontrolled polishing.
FIG. 19 is a graph showing the transition of the monitoring signal in the controlled polishing.
FIG. 20 is a graph for explaining predictive fuzzy control.
FIG. 21 is a schematic diagram for explaining predictive control.
FIG. 22 is a table showing an example of a fuzzy rule for predictive control.
FIG. 23 is a table showing another example of the fuzzy rule for predictive control.
FIG. 24 is a diagram illustrating a change in the circular locus on the impedance coordinate plane when the gap (pad thickness) between the conductive film and the sensor coil is changed.

Hereinafter, embodiments of the present invention will be described in detail with reference to FIGS.
FIG. 1 is a schematic diagram showing an overall configuration of a polishing apparatus according to an embodiment of the present invention. As shown in FIG. 1, the polishing apparatus includes a polishing table 12 having a polishing pad 10 affixed on its upper surface, and a top ring 14 that holds a wafer that is an object to be polished and presses it against the upper surface of the polishing pad 10. ing. The upper surface of the polishing pad 10 constitutes a polishing surface that is in sliding contact with a wafer that is an object to be polished.
The polishing table 12 is connected to a motor (not shown) disposed below the polishing table 12 and is rotatable about its axis as indicated by an arrow. Further, a polishing liquid supply nozzle (not shown) is installed above the polishing table 12, and the polishing liquid is supplied onto the polishing pad 10 from the polishing liquid supply nozzle.
The top ring 14 is connected to a top ring shaft 18, and is connected to a motor and a lifting cylinder (not shown) via the top ring shaft 18. As a result, the top ring 14 can be raised and lowered and rotated about the top ring shaft 18. On the lower surface of the top ring 14, a wafer as an object to be polished is sucked and held by a vacuum or the like.
In the above-described configuration, the wafer held on the lower surface of the top ring 14 is pressed against the polishing pad 10 on the upper surface of the rotating polishing table 12. At this time, the polishing liquid is supplied onto the polishing pad 10 from the polishing liquid supply nozzle, and the wafer is polished in a state where the polishing liquid exists between the polishing surface (lower surface) of the wafer and the polishing pad 10.
FIG. 2 is a schematic view showing a cross section of the top ring shown in FIG. As shown in FIG. 2, the top ring 14 includes a substantially disc-shaped top ring main body 31 connected to the lower end of the top ring shaft 18 via a universal joint portion 30, and a retainer disposed below the top ring main body 31. And a ring 32. The top ring body 31 is made of a material having high strength and rigidity such as metal or ceramics. The retainer ring 32 is made of a highly rigid resin material or ceramics. Note that the retainer ring 32 may be formed integrally with the top ring main body 31.
In a space formed inside the top ring main body 31 and the retainer ring 32, there is an elastic pad 33 that contacts the wafer W, an annular pressure sheet 34 made of an elastic film, and a substantially disk shape that holds the elastic pad 33. The chucking plate 35 is accommodated. The upper peripheral end of the elastic pad 33 is held by a chucking plate 35, and four pressure chambers (airbags) P1, P2, P3, and P4 are provided between the elastic pad 33 and the chucking plate 35. Yes. These pressure chambers P1, P2, P3, and P4 are supplied with pressurized fluid such as pressurized air through fluid passages 37, 38, 39, and 40, respectively, or are evacuated. The central pressure chamber P1 is circular, and the other pressure chambers P2, P3, P4 are annular. These pressure chambers P1, P2, P3, and P4 are arranged concentrically.
The internal pressures of the pressure chambers P1, P2, P3, and P4 can be changed independently from each other by a pressure adjusting unit (not shown), whereby four regions of the wafer W, that is, the central portion C1 and the inner intermediate portion. The pressing force on C2, the outer intermediate portion C3, and the peripheral portion C4 can be adjusted almost independently (of course, more precisely, the pressure chamber is affected to some extent by other regions such as adjacent regions). Further, by raising and lowering the entire top ring 14, the retainer ring 32 can be pressed against the polishing pad 10 with a predetermined pressing force. A pressure chamber P5 is formed between the chucking plate 35 and the top ring main body 31, and a pressurized fluid is supplied to the pressure chamber P5 via the fluid passage 41, or a vacuum is drawn. Yes. As a result, the entire chucking plate 35 and the elastic pad 33 can move in the vertical direction. A retainer ring 32 is provided around the wafer W to prevent the wafer W from jumping out of the top ring 14 during polishing.
As shown in FIG. 1, a sensor 50 for monitoring (detecting) the film state of the wafer W is embedded in the polishing table 12. This sensor 50 is connected to a monitoring device 53, and this monitoring device 53 is connected to a CMP controller 54. An eddy current sensor can be used as the sensor 50. The output signal of the sensor 50 is sent to the monitoring device 53, and the monitoring device 53 performs necessary conversion / processing (calculation processing) on the output signal (sensing signal) of the sensor 50 to generate a monitoring signal.
The monitoring device 53 also functions as a control unit that operates the internal pressure of each pressure chamber P1, P2, P3, P4 based on the monitoring signal. That is, the monitoring device 53 determines the force with which the top ring 14 presses the wafer W based on the monitoring signal, and transmits this pressing force to the CMP controller 54. The CMP controller 54 issues a command to a pressure adjusting unit (not shown) so as to change the pressing force of the top ring 14 against the wafer W. The monitoring device 53 and the control unit may be separate devices, or the monitoring device 53 and the CMP controller 54 may be integrated into a single control device.
FIG. 3 is a plan view showing the relationship between the polishing table 12 and the wafer W. As shown in FIG. As shown in FIG. 3, the sensor 50 has a center C of the wafer W being polished held by the top ring 14. _W It is installed at a position that passes through. Code C _T Is the center of rotation of the polishing table 12. For example, while the sensor 50 passes under the wafer W, the sensor 50 continuously increases in accordance with the film thickness of the conductive film such as a Cu layer of the wafer W or a change in film thickness on the trajectory (scanning line). Alternatively, a decreasing amount can be detected.
FIG. 4 shows a trajectory that the sensor 50 scans on the wafer W. That is, the sensor 50 scans the surface (surface to be polished) of the wafer W every time the polishing table 12 rotates, but when the polishing table 12 rotates, the sensor 50 generally has a center C of the wafer W. _W The trajectory passing through (the center of the top ring shaft 18) is drawn, and the surface to be polished of the wafer W is scanned. Since the rotation speed of the top ring 14 and the rotation speed of the polishing table 12 are usually different, the trajectory of the sensor 50 on the surface of the wafer W is the scanning line SL as the polishing table 12 rotates as shown in FIG. ₁ , SL ₂ , SL ₃ , ... and changes. Even in this case, as described above, the sensor 50 has the center C of the wafer W. _W The locus drawn by the sensor 50 is the center C of the wafer W every time. _W Pass through. In this embodiment, the timing of measurement by the sensor 50 is adjusted, and the center C of the wafer W is adjusted by the sensor 50. _W Is always measured every time.
The film thickness profile after polishing of the wafer W is the center C of the wafer W. _W It is known that the axis is generally subject to an axis that passes through and is perpendicular to the surface. Therefore, as shown in FIG. 4, the mth scanning line SL _m MP above the nth measurement point _mn The nth measurement point MP in each scanning line _1-n , MP _2-n , ..., MP _mn By tracking the monitoring signal for, the transition of the film thickness of the wafer W at the radial position of the nth measurement point can be monitored.
In FIG. 4, the number of measurement points in one scan is 15 for simplification. However, the number of measurement points is not limited to this, and can be various values depending on the measurement cycle and the rotation speed of the polishing table 12. When an eddy current sensor is used as the sensor 50, there are usually 100 or more measurement points on one scanning line. When the number of measurement points is increased in this way, one of the measurement points becomes the center C of the wafer W. _W The center C of the wafer W described above _W It is not necessary to adjust the measurement timing for.
FIG. 5 is a plan view showing an example of selecting measurement points to be monitored by the monitoring device 53 from the measurement points on the wafer W shown in FIG. In the example shown in FIG. 5, measurement points MP at positions corresponding to the vicinity of the center and the vicinity of the boundary lines of the regions C1, C2, C3, and C4 where the pressing forces are independently operated. _m-1 , MP _m-2 , MP _m-3 , MP _m-4 , MP _m-5 , MP _m-6 , MP _m-8 , MP _m-10 , MP _m-11 , MP _m-12 , MP _m-13 , MP _m-14 , MP _m-15 Monitoring. Here, unlike the example shown in FIG. _mi And MP _{m- (i + 1)} There may be another measurement point between and. Note that the selection of measurement points to be monitored is not limited to the example shown in FIG. 5, a point to be controlled on the surface to be polished of the wafer W can be selected as a measurement point to be monitored, and all the points on the scanning line can be selected. It is also possible to select a measurement point.
The monitoring device 53 performs a predetermined calculation process on the output signal (sensing signal) of the sensor 50 at the selected measurement point to generate a monitoring signal. Further, the monitoring device 53, based on the generated monitoring signal and a reference signal described later, corresponds to the pressure chambers P1, P2, P3 in the top ring 14 corresponding to the regions C1, C2, C3, C4 of the wafer W. , P4 pressures are calculated. That is, the monitoring device 53 compares the monitoring signal acquired for the measurement point selected as described above with the reference signal set for each measurement point in advance, and each monitoring signal converges to the respective reference signal. Therefore, the optimum pressure value of the pressure chambers P1, P2, P3, and P4 is calculated. The calculated pressure value is transmitted from the monitoring device 53 to the CMP controller 54, and the CMP controller 54 changes the pressure in the pressure chambers P1, P2, P3, and P4. In this way, the pressing force for each region C1, C2, C3, C4 of the wafer W is adjusted.
Here, in order to eliminate the influence of noise and smooth the data, a signal obtained by averaging monitoring signals for nearby measurement points may be used. Alternatively, the center of the surface of the wafer W is C _W Is divided into a plurality of regions concentrically according to the radius from, and the average value or representative value of the monitoring signal for the measurement points in each region is obtained, and this average value or representative value is used as a new monitoring signal for control. It may be used. Here, at each time point during polishing, C at each measurement point _W If the distance between the top ring 14 and the top ring 14 is determined, the top ring 14 is arranged in the radial direction of the polishing table 12 or when the top ring 14 is being polished. It is also possible to effectively cope with the case of swinging around the center. Since each measurement point actually has an area corresponding to the effective diameter measurement range of the sensor, the monitoring signal represents the state of a plurality of regions on the substrate in all the above cases. I can say that.
FIG. 6 is a diagram illustrating an effective measurement range of the sensor at each measurement point. When an eddy current sensor is used as the sensor 50, the effective measurement range on the wafer W is determined according to the size of the coil in the sensor 50, the effective range spread angle, and the distance from the sensor 50 to the wafer W. And the sensor 50 will acquire the information in the range shown with the dotted line of FIG. 6 in each measurement point. However, when the state of the peripheral edge of the wafer W is to be measured, a part of the effective measurement range of the sensor 50 protrudes from the surface to be polished of the wafer W (measurement point MP in FIG. 6). _m-1 , MP _m-N reference). In such a case, as shown in FIG. _m-1 , MP _m-N The monitoring signal corresponding to is greatly different from the monitoring signal in other regions, and the film thickness cannot be properly evaluated. A similar thing may occur depending on conditions for sensors other than eddy current sensors. In FIG. 7, the point at which the respective monitoring signals change from decreasing to constant at the end of the polishing time represents the polishing end point (at the time when the metal film is completely removed).
Therefore, in this embodiment, as a method for dealing with the problem that the monitoring signal value varies depending on the region on the wafer W even with the same film thickness, a reference signal is set for each of the regions C1 to C4 of the wafer W. This reference signal is a value (reference value) that serves as an index of the monitoring signal at each time point during polishing (polishing time point) in order to realize a desired film thickness profile (for example, a profile with a uniform film thickness after polishing). It can be expressed as a graph showing the relationship between the polishing time and the value of the desired monitoring signal at the polishing time. In this embodiment, a wafer of the same type as the wafer to be polished (hereinafter referred to as a reference wafer) is polished in advance, and each region C1 to C4 distributed in the radial direction of the wafer W based on the monitoring signal at this time. Create a reference signal for.
Here, if the pressing force for each region is operated in order to make the film thickness in the radial direction of the wafer uniform, the reference signal set for each region corresponds to the same film thickness at the same time. Must be a thing. In other words, if a reference signal that can be considered to correspond to the same film thickness at the same time is prepared for each region, and if the pressing force is operated so that the monitoring signal acquired for each region converges to these reference signals, each region The wafer W can be polished so that the film thickness at is uniform.
FIG. 8 is a flowchart showing a flow of polishing a reference wafer and creating a reference signal for each region based on the monitoring signal at this time. First, a wafer of the same type as the wafer to be polished, that is, a reference wafer is prepared. Then, as shown in FIG. 8, the film thickness distribution in the radial direction of the reference wafer before polishing is measured, and the representative film thickness before polishing of each of the regions C1 to C4 is acquired (step 1). Here, the same type of wafer means that the polishing rate (Removal Rate) at each time point during polishing is substantially equal to the wafer to be polished, and the monitoring signal acquired when the film thickness is the same is approximately equal to the wafer to be polished, and The film forming range at the wafer peripheral portion is substantially equal to the wafer to be polished. For example, in the eddy current sensor, the material of the film to be polished (metal film) of the reference wafer must be basically the same type as the film to be polished of the wafer to be polished. Also, if the resistance of the wafer to be polished is so small that it cannot be ignored compared to the resistance of the metal film and affects the monitoring signal during polishing, the resistance of the reference wafer should be approximately equal to the resistance of the wafer to be polished. I must. However, the reference wafer does not necessarily have to be exactly the same specification as the wafer to be polished. For example, when the polishing rate of the reference wafer is significantly different from the polishing rate of the wafer to be polished, the apparent polishing is performed by scaling (stretching or reducing) the monitoring signal when the reference wafer is polished with respect to the time axis. The speed can be adjusted and used for controlling the wafer to be polished. In order to allow sufficient control time, the initial film thickness of the reference wafer is preferably equal to or larger than the initial film thickness of the wafer to be polished. Even when the thickness is smaller, the polishing control is possible only by shortening the control time described later.
After obtaining the film thickness distribution of the reference wafer, the reference wafer is polished, and monitoring signals in the regions C1 to C4 are obtained (step 2). While the reference wafer is being polished, the pressures in the pressure chambers P1, P2, P3, and P4 with respect to the regions C1 to C4 are constant (invariable). However, the pressures in the pressure chambers P1, P2, P3, and P4 do not need to be equal to each other. Further, during polishing of the reference wafer, other polishing conditions such as the polishing pad 10, the polishing liquid, the rotation speed of the polishing table 12, and the rotation speed of the top ring 14 are basically constant. Preferably, the polishing conditions for polishing the reference wafer are the same as or similar to those for polishing the wafer to be polished.
After a predetermined time has elapsed, the polishing of the reference wafer is terminated. And the film thickness of the to-be-polished film on the reference | standard wafer after grinding | polishing is measured, and the representative film thickness after grinding | polishing of each area | region C1-C4 is acquired (step 3). When the film to be polished is a metal film, the polishing is stopped before the metal film is removed. This is because the measurement of the film thickness after polishing by the sensor 50 is assured and when the metal film is removed, the polishing rate changes greatly, and an accurate reference signal cannot be obtained. However, it is also possible to obtain the time point at which the metal film in each region of the wafer W is removed from the monitoring signal, and to create a reference signal with the film thickness at this time being 0. In this case, the metal film is completely Polish the reference wafer until it is removed.
As will be described later, in this embodiment, processing such as scaling and parallel movement is performed on the acquired monitoring signal of each region of the reference wafer to create a reference signal that allows the film thickness of each region to be uniform at each time point. Therefore, the film thickness is not necessarily uniform in polishing the reference wafer. However, since there is a problem in grasping the steep film thickness profile by the sensor, it can be expected that a more accurate reference signal is obtained as the film thickness in the reference wafer radial direction before and after polishing is uniform.
In general, when there is local unevenness in the film thickness profile of the wafer, if this unevenness is smaller than the effective measurement range of the sensor, the sensor cannot output a signal that accurately reflects the shape of the unevenness. For example, as shown in FIG. 9A, it is assumed that there is a steep convex portion at point a on the wafer. Since the effective measurement range of the sensor has a certain size, the sensor does not output a value corresponding to the peak film thickness of this convex part, but a signal corresponding to the film thickness averaged within the effective measurement range. A value will be output. Therefore, in the film thickness measurement before and after polishing the reference wafer, it is preferable to average the measurement values acquired in the region corresponding to the effective measurement range of the sensor 50 to obtain the film thickness value at the center point of this region. The film thickness distribution obtained in this manner is shown in FIG. 9B. 9A and 9B, black points on the graphs indicate measurement points of the sensor 50.
Next, in steps 4 and 5 (see FIG. 8), the reference signal is corrected so that each reference signal can be regarded as corresponding to the same film thickness at the same time.
FIG. 10 is an example of a monitoring signal when the reference wafer is polished. In general, the value of the monitoring signal (and sensor signal) does not indicate the film thickness itself, but the value of the monitoring signal and the film thickness have a certain relationship. However, as described above, even if the film thickness is the same, the monitoring signal acquired at the periphery of the wafer W is smaller than the monitoring signal at the other region, and is obtained due to the influence of the conductive material or the like. May not show the value that should be originally obtained. Therefore, the film thickness before and after polishing measured in steps 1 and 3 is assigned to the monitoring signal, thereby associating the monitoring signal with the film thickness. Specifically, as shown in FIG. 10, the film thickness d before and after polishing of the reference region C0. _C0 S, d _C0 E is allocated to the start point and end point of the monitoring signal in the reference area C0. Similarly, the film thickness d before and after polishing of the area Ci other than the reference area _Ci S, d _Ci E is assigned to the start point and end point of the monitoring signal in the area Ci. As the reference region C0, for example, a region C1 including the center portion of the wafer can be selected.
FIG. 11 is a diagram for explaining the scaling with respect to the time axis of the monitoring signal. In step 4, the monitoring signal is scaled along the time axis so that the average polishing rates of the regions C1 to C4 are the same. Here, the scaling means that the monitoring signal is expanded or reduced along the time axis.
It is assumed that the polishing time when the reference wafer is previously polished is TE. At this time, the average polishing rate R in the reference region C0 is expressed by the following equation (1).
R = (d _C0 S-d _C0 E) / TE (1)
Therefore, the correction polishing time for the region Ci is set so that the average polishing rate of the region Ci is equal to the average polishing rate of the reference region C0.
T _i E = (d _Ci S-d _Ci E) / R (2)
far.
Then, the original polishing start time is set to 0, and the time t corresponding to each signal value of the monitoring signal in the area Ci. _i Is corrected as shown in the following equation (3).
t _i ← t _i × T _i E / TE (3)
In the above formula (3), the symbol “←” represents replacement.
In FIG. 11, d _Ci S-d _Ci E> d _C0 S-d _C0 An example in the case of E is shown.
FIG. 12 is a diagram for explaining a method of translating the monitoring signal scaled along the time axis in this way and further translating along the time axis, and is a diagram for explaining step 5 in FIG. In this step 5, an operation of aligning the initial film thickness in each region is performed.
Now, it is assumed that the polishing rate is approximately constant in each of the regions C1 to C4 at each time point during polishing of the reference wafer. At this time, the initial film thickness d in the reference region C0 _C0 S is the initial film thickness d in the area Ci _Ci Polishing time Δt required for polishing until S matches _i Is obtained from the following equation (4).
Δt _i = (D _C0 S-d _Ci S) / R (4)
Therefore, the polishing time t in the area Ci corrected by the above equation (3). _i Is further corrected using the following equation (5).
t _i ← t _i + Δt _i ... (5)
In the example shown in FIG. 12, each signal value of the monitoring signal in the area Ci is expressed by Δt along the time axis. _i If the movement is performed only by the parallel movement, the film thickness of the area C0 and the film thickness of the area Ci at the time TiS, which is the start point of the monitoring signal of the area Ci, can be regarded as being equal to each other. Furthermore, since the average polishing rate is the same between the region C0 and the region Ci, the film thicknesses at the time TE can be regarded as being equal to each other. Therefore, the film thickness d at time TX (where TiS ≦ TX ≦ TE) _C0 X and d _Ci X can be considered equal to each other.
As described above, since the monitoring signal of the area Ci is scaled along the time axis and further translated, the monitoring signal of the area C0 and the monitoring signal after correction of the area Ci are generally from Max (0, TiS). Both exist only in the section up to Min (TE, TiE + Δti). Here, Max indicates a larger value in parentheses, and Min indicates a smaller value. Figure 12 shows d _C0 S> d _Ci An example of S is shown, of course d _C0 S <d _Ci In some cases, the polishing start time TiS of the region Ci is a negative value.
Next, the waveform of the reference signal of each region obtained in this way is smoothed as necessary to reduce the noise component (step 6). As a smoothing method, a moving average, a more general digital filter, or polynomial regression can be applied. And the process of the above-mentioned step 4-6 is repeated, and the reference signal about all the area | regions C1-C4 is defined. At this stage, the time for each signal value of the reference signal is corrected independently for each region and generally takes a different value. Therefore, the reference signal for each region is interpolated and the time for the same time at a fixed time interval is obtained. It is also possible to redefine the reference signal.
As can be seen from FIG. 12 and Equation (4), the initial film thickness d _Ci As S is smaller, the starting point TiS of the reference signal moves to the right in FIG. The final film thickness d _Ci As E increases, the end point TiE + Δti of the reference signal moves to the left in FIG. Since the initial film thickness and the final film thickness generally vary depending on the region, when the reference signal is obtained for each region, the start point and the end point of each reference signal usually do not match. Therefore, the reference signal is set as follows. First, the initial film thickness in each region is compared with each other, and the minimum value Min (d _Ci S). Similarly, the final film thickness in each region is compared with each other, and the maximum value Max (d _Ci E). And Min (d _Ci From the time corresponding to S), Max (d _Ci Only the monitoring signal in the section up to the time corresponding to E) is used as the reference signal. Alternatively, it is also possible to define a reference signal in a wider section by extrapolating the monitoring signal in each region so that the control time can be extended.
The reference signal for each area obtained in this way is stored in a storage unit (for example, a hard disk) of the monitoring device 53. When polishing the wafer W, the pressing force of the pressure chambers P1, P2, P3, and P4 on the wafer W is manipulated so that the monitoring signals in the regions C1 to C4 converge on the reference signal. In addition, although the example which sets a reference signal regarding the area | regions C1-C4 corresponding to the pressure chambers P1-P4 was demonstrated above, as above-mentioned, since a monitoring signal can be produced | generated with respect to not only this but various areas, The reference signal can be defined not only for the regions C1 to C4 but also for various regions on the surface of the wafer W.
According to the above-described embodiment, since the reference signals indicating the same film thickness are obtained at the same time, the pressure chambers P1, P2, P3 and the like so that the monitoring signals acquired in the respective regions converge on the respective reference signals. By manipulating the pressure of P4, it is possible to polish with a uniform film thickness. Therefore, as shown in FIG. 7, even when the monitoring signal at the peripheral edge of the wafer W is extremely small compared to other regions, a uniform final film thickness can be obtained. In addition, since the reference signal is defined for each region, a desired non-uniform residual film thickness profile is realized by appropriately translating each reference signal created as described above with respect to the time axis. You can also.
For example, the remaining film thickness of the area Ci is larger than the area C0 by Δd _Ci When it is desired to realize a film thickness profile that is as large as _i After the correction, the polishing time t _i Is corrected using the following equation (5) ′.
t _i ← t _i + Δd _Ci / R (5) '
In other words, the polishing time t is calculated by using the following equation (4) ′ instead of the equation (4). _i Correct.
Δt _i = (D _C0 S-d _Ci S + Δd _Ci ) / R (4) '
Where Δd _Ci If <0, the remaining film thickness of the region Ci is -Δd than that of the region C0. _Ci Will only be small.
In this way, in FIG. 12, the film thickness of the area Ci at the time TiS, which is the starting point of the monitoring signal of the area Ci, is Δd from the film thickness of the area C0. _Ci Can only be considered big. Further, since the average polishing rate is the same between the region C0 and the region Ci, the film thickness of the region Ci and the film thickness of the region C0 at an arbitrary time TX (TiS ≦ TX ≦ TE in the example of FIG. 12). From Δd _Ci Can only be considered big. Therefore, if the monitoring signal of each region created in this way is used as a reference signal, and the pressing force is manipulated so that the monitoring signal acquired for each region during polishing converges to these reference signals, The thickness of the region Ci is larger than the thickness of the region C0 by Δd _Ci It is expected to achieve a desired profile that is only as large as possible.
In this way, for example, when the uppermost layer is a metal film and there is an insulating layer below that, and there is a wiring, by defining the target profile of the remaining thickness of the metal film by knowing the distribution of the thickness of the insulating layer, Polishing can be advanced so that the height from the wiring is uniform. In the following, detailed description will be made focusing on the case where the profile of the remaining film thickness of the film to be polished is made uniform.
FIG. 13 shows a monitoring signal MS for a certain area on the wafer. ₁ For the reference signal RS set for this ₀ And a new monitoring signal MS based on the straight line B ₂ It is the graph which showed the method of converting into. Here, the straight line B is the reference signal RS. ₀ It is a straight line with an inclination of -1 passing through the polishing end point. For example, as shown in FIG. ₁ Monitoring signal MS ₁ Value of v ₁ Is given, the reference signal RS ₀ A point P having the same value is obtained. Then, from the time of this point P, the reference signal RS ₀ The remaining time T until the polishing end point is obtained. The remaining time T can be obtained by referring to the straight line B as can be seen from FIG. New monitoring signal MS based on required time T ₂ Time t ₁ Signal value at ₂ Set. For example, v ₂ = Signal value v such that T ₂ Set. Alternatively, the signal value v ₂ The time T from the start of polishing to the end of polishing in the reference signal ₀ Normalize with v ₂ = T / T ₀ At this time, the straight line B takes the value 1 at time 0 and the slope is -1 / T. ₀ It becomes a straight line.
Reference signal RS ₀ If the same concept is applied to the above, the above-described straight line B can be regarded as a new reference signal for the converted monitoring signal. This new reference signal (straight line B) is the reference signal RS. ₀ Since it represents the remaining time from each of the above points to the polishing end point, it becomes a linear monotonously decreasing function with respect to time, and the control calculation becomes easy.
In the case of controlling with the aim of a profile with a uniform film thickness after polishing, if the same conversion is performed on the monitoring signal of each region on the wafer W using the set reference signal, the converted monitoring signal Is expressed as a remaining time until the polishing end point in the corresponding reference signal or a value obtained by normalizing the remaining time. However, since the reference signals can be regarded as corresponding to the same film thickness at the same time, the monitoring signals of all the regions can be simply compared with each other as an index indicating the film thickness. At this time, all the converted reference signals coincide with the straight line B and are unified.
Further, in this case, in many cases, a new monitoring signal MS after conversion is obtained. ₂ Varies linearly in proportion to the film thickness of the polished surface of the wafer. Therefore, even when the film thickness value of the surface to be polished cannot be measured due to the influence of the polishing liquid, the wiring pattern on the surface to be polished of the wafer, the lower layer, etc., it is possible to obtain good control performance by linear calculation. In the example shown in FIG. 13, the reference signal RS ₀ Although the polishing end point in FIG. 1 was described as the reference time, the reference signal RS ₀ The reference time in is not limited to the polishing end point. For example, the reference signal RS ₀ A reference time can be arbitrarily determined such as a time at which takes a predetermined value. In particular, as described above, when polishing is controlled so that the remaining film thickness has a non-uniform profile, all regions in the reference signal do not reach the polishing end point at the same time, and therefore parallel according to equation (4) ′. One point on the time axis is determined as a common reference time for the reference signal of each area created by movement. The converted reference signals at this time are all unified with the straight line B as in the case of the uniform profile. Note that, in a section where the reference signal value corresponding to the monitoring signal value does not originally exist or a section where the monitoring signal value does not change with the polishing time, the value of the new monitoring signal after conversion is indefinite. In such a case, the control is stopped, and the conventional value may be maintained as the set value such as the pressing force of the top ring. In FIG. 13, the reference signal exists until the polishing end point is reached. This is because the reference wafer is polished until the polishing end point is passed, the polishing end point is detected based on the monitoring signal, and the reference signal is defined with the film thickness at this time being zero.
FIG. 14 is a graph showing an example of a method of applying the reference signal converted as described above. In FIG. 14, the reference signal RS is set so that the polishing time up to the polishing end point becomes a desired value at the polishing start point or control start point. ₁ Is translated along the time axis to create a new reference signal RS ₂ Is set. Incidentally, at the polishing start time or control start time, the reference signal RS ₁ If the polishing time to the polishing end point is a desired value, the reference signal RS ₁ The amount of parallel movement may be zero.
Then, the reference signal RS with respect to the time axis ₂ Monitoring signal MS _A , MS _B , MS _C And monitoring signals in other areas not shown are the reference signal RS. ₂ Control is performed so as to converge. In this way, even if the value of the monitoring signal before conversion in a certain region is different from other regions at the same film thickness, not only the in-plane uniformity can be improved regardless of the initial film thickness profile. Even if the initial film thickness varies between wafers or the state of an apparatus such as a polishing pad changes, it can be expected that the time until the polishing end point becomes a predetermined value. Thus, if the polishing time can be made constant, the wafer can be transported in the polishing apparatus at a substantially constant cycle that can be expected. Therefore, the throughput is improved without being delayed by the wafer having a long polishing time.
FIG. 15 is a graph showing still another example of the reference signal application method. In FIG. 15, the reference signal RS is set so that the value av obtained by averaging the monitoring signal values in each region coincides with the reference signal at the polishing start time or control start time. ₃ Is translated along the time axis to create a new reference signal RS ₄ Set. Here, the monitoring signal value averaging method may be any method as long as it obtains a value representative of the progress of wafer polishing. For example, an arithmetic average or a weighted average is calculated. Or a method of taking a median value.
Then, the reference signal RS with respect to the time axis ₄ Monitoring signal MS _A , MS _B , MS _C And other area monitoring signals not shown are the reference signal RS. ₄ Control is performed so as to converge. In this way, even if the value of the monitoring signal in a certain region is different from that in the other regions at the same film thickness, the pressing force on each region C1 to C4 of the wafer W can be compared with the example shown in FIG. It is not necessary to change the operation amount extremely, and stable polishing can be expected. In addition, it is expected that the polishing time after starting polishing or after starting control will be equal to the polishing time when polishing from the same film thickness when acquiring the reference signal, improving the in-plane uniformity regardless of the initial film thickness profile In addition, the average polishing rate can be realized regardless of the state of the polishing pad or the like.
FIG. 16 is a graph showing still another example of the reference signal application method. In FIG. 16, a value obtained by averaging the monitoring signals in each region at a predetermined cycle is a reference signal RS. ₅ To match the reference signal RS ₅ Is translated along the time axis. For example, the value av obtained by averaging the monitoring signal ₁ , Av ₂ , Av ₃ To match the reference signal RS ₅ Respectively, and a new reference signal RS ₆ , RS ₇ , RS ₈ Set each. Then, the pressing force or the like for each of the regions C1 to C4 of the wafer is manipulated so that the monitoring signal of each region converges to a reference signal set by parallel movement every moment. In this way, even if the value of the monitoring signal in a certain region is different from the other regions at the same film thickness, the pressing force of each region C1 to C4 of the initial wafer is generally within a reasonable range. If the pressing force in a certain region increases, the pressing force in another region decreases. Therefore, although this embodiment does not have a function of adjusting the polishing time and the polishing rate, stable polishing can be performed by reducing the change in the operation amount. Furthermore, excellent in-plane uniformity can be achieved regardless of the initial film thickness profile.
In such a case, a good result can be obtained even if the polishing of the pattern wafer is controlled using a reference signal created using a blanket wafer as a reference wafer. Here, the blanket wafer refers to a wafer in which one or more materials are formed on the wafer with a uniform thickness, and a so-called pattern is not formed. In general, in polishing a pattern wafer, the polishing rate is different from that of a blanket wafer, and before and after the unevenness of the surface to be polished is eliminated. Further, if the film to be polished is a metal film and the sensor is an eddy current sensor, the rate of change of the monitoring signal with respect to the film thickness varies before and after the surface irregularities are eliminated. However, since the film thickness profile is controlled by the above method and there is no function to adjust the polishing rate, good control performance is expected regardless of the difference in the polishing rate and the change rate of the monitoring signal. it can.
It is difficult to measure the thickness of a pattern wafer if the film thickness is small, and it is only complicated to create a reference signal by polishing it in advance each time the type of product wafer to be polished changes. Instead, the product wafer is wasted. Therefore, it is practically significant to be able to control the polishing of the pattern wafer by applying the reference signal from the blanket wafer.
15 and FIG. 16, the example in which the reference signal is translated so as to coincide with the value obtained by averaging the monitoring signal at the start of polishing or at a predetermined period has been described. However, a value other than the value obtained by averaging the monitoring signal is used. It is also possible to translate the reference signal as a reference. For example, the reference signal may be translated based on the monitoring signal of a predetermined area of the wafer. That is, at the start of polishing, the reference signal may be translated so that the reference signal matches the monitoring signal of the predetermined area at the start of polishing, and even during the polishing process, the reference signal is the predetermined area at that time. The reference signal may be translated so as to coincide with the monitoring signal.
As described above, if the reference signal is appropriately defined for the wafer to be polished, the reference signal is defined, and the pressing force is operated based on the reference signal, then monitoring each part of the wafer every moment during polishing. The film thickness profile can be easily controlled without the complicated operation of individually determining the relationship between the signal and the film thickness.
FIG. 17 is a graph showing the radial film thickness distribution before and after polishing in the case where polishing is performed by creating a reference signal by the method of this embodiment, aiming at a uniform film thickness profile after polishing. . In the controlled polishing (polishing method of the present embodiment), the pressing force was manipulated so that the monitoring signal for each region converged to each reference signal. On the other hand, in uncontrolled polishing, a pressing force equal to the initial pressing force at the time of controlled polishing was applied to the wafer at a constant level. From FIG. 17, it can be seen that good residual film thickness uniformity including the peripheral edge of the wafer can be obtained.
FIG. 18 is a graph showing the transition of the monitoring signal in uncontrolled polishing, and FIG. 19 is a graph showing the transition of the monitoring signal in controlled polishing. As shown in FIG. 18, in the uncontrolled polishing, the values of the monitoring signals in the three regions on the wafer surface (center portion, inner intermediate portion, and outer intermediate portion) are different. On the other hand, in the controlled polishing, as shown in FIG. 19, it can be seen that the monitoring signal is generally converged to one value. Concerning the peripheral portion of the wafer, since the monitoring signal value is greatly separated from other regions for the reason described above, the convergence cannot be visually confirmed from the figure. However, in practice, since polishing control is performed along the corrected reference signal also at the peripheral portion of the wafer, a uniform film thickness is obtained in all regions including the peripheral portion as shown in FIG. .
FIG. 20 is a graph for explaining an example of the control calculation method according to the present invention. In FIG. 20, the monitoring signal conversion method described with reference to FIG. 13 is used. A new reference signal ys (t) at time t after the start of polishing is expressed by the following equation (X).
ys (t) = T ₀ -T (6)
In the above formula (6), T ₀ Is the time from the start of polishing to the end of polishing in the reference signal.
Where T ₀ Is for the reference signal translated with respect to the time axis by any one of the two previous methods out of the three methods described above (see FIGS. 14 and 15). In the example shown in FIG. 16, the right side is a value obtained by averaging the monitoring signals of the respective regions at that time. At this time, t _o From t to t _o Predicted value y of monitoring signal in each area of wafer after elapse _p (T, t _o ) Is represented by the following formula (7).
y _p (T, t _o )
= Y (t) + t _o {Y (t) -y (t-Δt _m )} / Δt _m ... (7)
In the above equation (7), y (t) is a monitoring signal at time t, Δt _m Is a time determined for calculating the inclination of the monitoring signal with respect to time.
At this time, from time t to t _o The discrepancy D (t, t) of the predicted value of the monitoring signal after the lapse of time with respect to the reference signal _o ) Is defined as the following equation (8).
D (t, t _o ) =-{Y _p (T, t _o -Y _s (T + t _o )} / T _o ... (8)
If the degree of mismatch D represented by the equation (8) is positive, it means that the monitoring signal is advanced with respect to the reference signal, and if it is negative, it means that the monitoring signal is delayed.
As shown in FIG. 20, when the reference signal is a straight line, if the predicted value of the monitoring signal always coincides with the reference signal at each time point of the period Δt, the monitoring signal is expected to asymptotically converge on the reference signal. The Therefore, for example, as shown in FIG. 21, the unmatching degree of the region C3 of the wafer to which the pressing force u3 is applied to the back surface is D3, and the unmatching degrees of the regions C2 and C4 adjacent to the region C3 are D2 and D4, respectively. Consider the determination of the change amount Δu3. FIG. 22 is an example of a fuzzy rule for determining such a change amount Δu3 of the pressing force u3. FIG. 23 shows the temperature T of the polishing pad immediately after sliding on the wafer in addition to the fuzzy rule of FIG. _p It is an example of a fuzzy rule when 22 and FIG. 23, “S” is “small”, “B” is “large”, “PB” is “largely increased”, “PS” is “smallly increased”, “ZR” is “not changed”, “NS” means “slightly reduce” and “NB” means “reduced greatly”.
As shown in the fuzzy rule of FIG. 22, the change amount Δu3 of the pressure increases as the mismatch degree D3 of the corresponding area C3 or the pressing force u3 itself decreases, and the mismatch between the areas C2 and C4 adjacent to the area C3. Even when the degrees D2 and D4 are small, the adjustment is made to increase. If the fuzzy rules are determined in the same way for the pressing force of other areas that are independent from each other, the degree of inconsistency of the corresponding areas, and the amount of change in pressing force, the pressing force is extremely large or small. Control can be performed so that all the inconsistencies converge to zero without changing to values.
Further, in the example shown in FIG. 23, in many cases, the polishing pad temperature T is increased in consideration of the fact that the polishing rate increases as the polishing pad temperature increases, and thus the temperature easily increases. _p Is lower, the change amount Δu3 of the pressing force u3 is increased, and the temperature T _p The higher the is, the smaller the change amount Δu3 is set.
The applicable fuzzy rules are not limited to those shown in FIGS. 22 and 23, and can be arbitrarily defined according to the characteristics of the system. In addition, inference methods such as membership functions, logical product methods, implication methods, accumulation methods, and defuzzification methods for the antecedent and consequent variables can be selected and used as appropriate. For example, if the membership function of the consequent part is set appropriately, the change amount Δu3 of the pressing force can be adjusted, and upper and lower limit constraints are further imposed on the pressing force u3 and the change amount Δu3 thus obtained. Can also be determined. Further, the region for defining the monitoring signal or the degree of inconsistency is not limited to the above-described C1 to C4. For example, one or two regions are added to the boundary portion to make the region finer. It is also possible to perform control.
In the above example, if the original reference signal and the monitoring signal are nearly linear with respect to time, the conversion of the monitoring signal described with reference to FIG. 13 to the value related to the polishing time is not necessarily required. When the time change of the monitoring signal is represented on the graph and the curvature is small, the time t obtained by the equation (7) is obtained in the same manner as in FIG. _o If the predicted value of the subsequent monitoring signal always coincides with the reference signal ys (t), it is expected that the monitoring signal gradually approaches the reference signal and good control is performed. When the monitoring signal is not converted into a value related to time, in the parallel movement of the reference signal described with reference to FIG. 15 or FIG. 16, for example, an area including a wafer peripheral part or an area where the monitoring signal differs greatly due to the influence of SUS parts is excluded. Thus, an average value that is a reference for the parallel movement may be obtained.
In the above-described example, the prediction type fuzzy control that obtains the prediction value of the inconsistency and performs the inference is used. After the sensor captures information on the surface to be polished of the wafer, until the actual pressing force is completely replaced with the new value and the polishing state changes, and the sensor output value changes completely, output from the sensor to the monitoring device Many steps are required, such as signal transfer, conversion and smoothing to a monitoring signal, calculation of pressing force, transfer to a control unit, command to a pressure adjusting unit, operation of a pressing mechanism (pressure chamber). Therefore, it usually takes about 1 to 2 seconds until the change in the operation amount is completely reflected in the signal waveform. Thus, in order to perform effective control while suppressing the influence of response delay, predictive control is effective.
As a predictive control method, not only the above-described fuzzy control but also, for example, an appropriate mathematical model may be defined to perform model predictive control. If modeling is performed including the above-described response delay, further improvement in control performance can be expected. In such a system, even if the control cycle is shortened, the next operation is performed before the change in the operation amount is sufficiently reflected in the monitoring signal. There is a risk of causing an unnecessary change in the operation amount and a change in the signal value due to this. On the other hand, the polishing time is usually about several tens of seconds to several hundreds of seconds. Therefore, if the control period is too long, the polishing end point is reached before in-plane uniformity is achieved. Therefore, the control cycle is preferably 1 second or more and 10 seconds or less.
When polishing the target wafer while operating the pressing force, the polishing end point (the point where the polishing conditions are switched) can be detected simultaneously from the monitoring signal when the metal film is removed or when a predetermined threshold value is reached. Can be detected.
Further, the reference signal as described above may be defined only for the two regions of the region C1 (the center portion of the wafer) and the region C4 (the peripheral portion of the wafer). In this case, the reference signal of the region C1 is used when controlling the region C1 and the regions C2 and C3 (inner intermediate portion and outer intermediate portion). Preferably, as described above, a reference signal may be defined for each region on the wafer surface, and the reference signal corresponding to each region may be used during polishing. This not only eliminates the influence of the monitoring signal change at the peripheral edge of the wafer, but also affects the monitoring signal from the eddy current sensor when a conductive or magnetic part such as a SUS flange is in the top ring. In the case of exerting good control performance, good control performance can be obtained by eliminating the influence.
In the process of defining the reference signal, the monitoring signal is scaled and translated while assuming that the polishing rate of each region is constant during polishing of the reference wafer, but the polishing time is sufficiently long. If the initial film thickness and polishing rate are not extremely different between regions, the amount of scaling and translation is small, and the practicality of grasping the film thickness profile by the monitoring signal is not impaired.
In the above-described embodiment, the case where the monitoring signal monotonously decreases with the progress of polishing has been described. However, the present invention can also be applied to the case where the monitoring signal monotonously increases. For example, when an impedance-type eddy current sensor is used as the sensor 50, the following method disclosed in Japanese Patent Laid-Open No. 2005-121616 can be applied.
As shown in FIG. 1, the conductive film present on the surface of the wafer W is measured via a polishing pad 10 from a sensor (eddy current sensor) 50 embedded in the polishing table 12. In this case, the gap between the sensor 50 and the conductive film changes according to the thickness of the polishing pad 10 interposed therebetween. As a result, for example, as shown in FIG. 24, the arc trajectories of the signal component X and the signal component Y vary according to the gap (gap) G corresponding to the thickness (t1 to t4) of the polishing pad 10 to be used. Therefore, in order to measure the film thickness of the conductive film of the semiconductor wafer W with high accuracy from the arc locus of the signal component X or the signal component Y, the thickness of the polishing pad to be used (before using the polishing pad) It is necessary to measure the film thickness of the conductive film to be measured after preparing measurement information of the signal component X and the signal component Y at a known film thickness.
However, from the measurement results of the signal component X and the signal component Y by the eddy current sensor, as shown in FIG. 24, the X component and the Y component regardless of the gap G between the sensor coil end and the conductive film. When the output values for each film thickness of the conductive film are connected by straight lines (r1 to r3), an intersection (center point) P at which the straight lines intersect can be acquired. This preliminary measurement straight line rn (n: 1, 2, 3,...) Corresponds to the thickness of the conductive film with respect to a reference line (horizontal line in FIG. 24) L where the signal component Y passing through the intersection P is constant. It tilts at an elevation angle θ.
Therefore, even if the thickness of the polishing pad for polishing the conductive film of the semiconductor wafer W is unknown, the measurement result (output value) and the center of the signal component X and the signal component Y of the conductive film to be polished If the elevation angle θ with respect to the reference line L of the measurement line rn connecting the points P is obtained, the conductivity of the measurement target is determined based on the correlation with the change tendency of the elevation angle θ according to the film thickness of the conductive film that has been preliminarily measured. The film thickness of the conductive film can be derived. However, in order to control the residual film thickness uniformity, it is not always necessary to know the film thickness absolute value, and it is only necessary to relatively grasp the film thickness in the radial direction of the wafer W. Therefore, the elevation angle θ is simply used as the monitoring signal. The reference line L may be a vertical line in FIG. 24 in which the reactance component X is constant.

  The present invention is applicable to a polishing apparatus and a polishing method for polishing and flattening a substrate such as a semiconductor wafer.

Claims (18)

A polishing apparatus for polishing a substrate having a film formed on a surface thereof,
A polishing table having a polishing surface;
A top ring that presses the substrate against the polishing table by independently applying a pressing force to the first plurality of regions on the substrate;
A sensor for detecting the state of the film at a plurality of measurement points;
A monitoring device that generates a monitoring signal for each of the second plurality of regions on the substrate from the output signal of the sensor;
A storage unit storing a plurality of reference signals indicating a relationship between a reference value of the monitoring signal and a polishing time;
A control unit that operates a pressing force on the first plurality of regions so that the monitoring signal corresponding to each of the second plurality of regions converges on any one of the plurality of reference signals. A polishing apparatus characterized by that. One of the second plurality of regions is a region including a peripheral portion of the substrate,
The polishing apparatus according to claim 1, wherein one of the plurality of reference signals is a reference signal for a region including a peripheral portion of the substrate.   The polishing apparatus according to claim 1, wherein the plurality of reference signals are provided corresponding to the second plurality of regions, respectively.   The signal value of the monitoring signal and the signal value of the reference signal are converted into a value related to a polishing time based on the reference signal to generate a new monitoring signal and a new reference signal. Item 4. The polishing apparatus according to any one of Items 1 to 3.   A value obtained by averaging the new monitoring signal in the second plurality of regions at an arbitrary time of the polishing step is obtained, and the new reference signal at the time coincides with the averaged value. The polishing apparatus according to claim 4, wherein the new reference signal after the time is translated with respect to the time axis.   The polishing apparatus according to claim 1, wherein the plurality of reference signals correspond to the same film thickness at the same time.   The plurality of reference signals correspond to film thicknesses reflecting a predetermined film thickness difference set between the second plurality of regions at the same time point. The polishing apparatus according to item.   The polishing apparatus according to claim 1, wherein a control period of the control unit is not less than 1 second and not more than 10 seconds.   The polishing apparatus according to claim 1, wherein the sensor is an eddy current sensor.   10. The polishing apparatus according to claim 1, wherein the control unit detects a polishing end point based on a monitoring signal generated by the monitoring apparatus. 11. A polishing method in which a substrate is pressed against a polishing table by applying an independent pressing force to the first plurality of regions on the substrate and polished.
Define multiple reference signals that indicate the relationship between the reference value of the monitoring signal related to the film thickness on the substrate and the polishing time,
The state of the film on the substrate at multiple measurement points is detected using a sensor,
Generating a monitoring signal for each of the second plurality of regions on the substrate from the output signal of the sensor;
Polishing characterized by operating a pressing force on the first plurality of regions so that the monitoring signal corresponding to each of the second plurality of regions converges to any one of the plurality of reference signals. Method. One of the second plurality of regions is a region including a peripheral portion of the substrate,
The polishing method according to claim 11, wherein one of the plurality of reference signals is a reference signal for a region including a peripheral portion of the substrate.   The polishing method according to claim 11, wherein the plurality of reference signals are provided corresponding to the second plurality of regions, respectively.   The polishing method according to claim 11, wherein the plurality of reference signals are obtained by polishing a blanket wafer.   The polishing method according to claim 11, wherein the plurality of reference signals correspond to the same film thickness at the same time. Prepare a reference substrate of the same type as the substrate to be polished,
Measure the thickness of the reference substrate,
Polishing the reference substrate and detecting the state of the film on the reference substrate at a plurality of measurement points by the sensor;
Generating a monitoring signal in the first region and the second region selected from the second plurality of regions from the output signal of the sensor;
The polishing is stopped when the film to be polished in the first region and the second region is completely removed,
Determining an average polishing rate of the first region and the second region;
Stretching or reducing the monitoring signal of the second region along the time axis so that the average polishing rate of the second region matches the average polishing rate of the first region;
Determining the polishing time required for the initial film thickness of the second region to match the initial film thickness of the first region;
The monitoring signal of the stretched or shrunk second region is translated along the time axis by the determined polishing time,
The polishing method according to claim 15, wherein the plurality of reference signals are defined by using the translated monitoring signal as a reference signal of the second region.   15. The plurality of reference signals correspond to film thicknesses that reflect a predetermined film thickness difference set between the second plurality of regions at the same time point. The polishing method according to item. Prepare a reference substrate of the same type as the substrate to be polished,
Measure the thickness of the reference substrate,
Polishing the reference substrate and detecting the state of the film on the reference substrate at a plurality of measurement points by the sensor;
Generating a monitoring signal in the first region and the second region selected from the second plurality of regions from the output signal of the sensor;
Measure the thickness of the reference substrate after polishing,
Determining an average polishing rate of the first region and the second region;
Stretching or reducing the monitoring signal of the second region along the time axis so that the average polishing rate of the second region matches the average polishing rate of the first region;
Determining the first polishing time required for the initial film thickness of the second region to match the initial film thickness of the first region;
A second polishing time required for the initial film thickness of the second region to have a predetermined film thickness difference from the initial film thickness of the first region;
The monitoring signal of the stretched or shrunk second region is translated along the time axis by the sum of the first polishing time and the second polishing time,
The polishing method according to claim 17, wherein the plurality of reference signals are defined by using the translated monitoring signal as a reference signal for the second region.

Images